Part- Of-Speech (Pos) Tagging of Limbu Language Using Artificial Neural Networks

Abstract

This paper aims to explore the use of neural networks for part-of-speech tagging (PoS) of Limbu language, a Tibeto-Burman language spoken in eastern Nepal, North-Eastern India, Bhutan etc. The Limbu language has a complex morphology with a rich system of inflections and derivations. Computational implementation for an under-resource language is strenuous for any natural language processing works. All through the process of part-of-speech tagging of Limbu language we face many impediments as our work is the initial work or the only towards Natural language Processing of Limbu language. There exists no computational work for the language as well all the pre-processing needed to be carried out by us including corpus generation. In this study we used a typical approach Feedforward Neural Network. The performance of the tagger is evaluated using various metrices such as accuracy, precision, recall and F1 score. We also compared the achievement of implemented model with other stochastic models like Hidden Markov Model (HMM). Our findings show that the neural network-based tagger achieves competitive results, outperforming stochastic models, and has the potential to be used in various natural language processing applications for Limbu language. The applied method imparted excellent accuracy for the corpus used. INDEX TERMS Limbu, Part-of-Speech (PoS) Tagging, Natural Language Processing (NLP), Hidden Markov Model (HMM), Feed Forward Neural Network, BiLSTM.